PARAMETERS OF THE CURVE OF LOCAL ESTIMATE DISTRIBUTION AS IMAGE QUALITY MEASURES

Abstract

This article focuses on selecting the best quality image from the series without the reference image. The results of studies of a new approach to digital image quality assessment, based on the local quality estimates distribution, are presented. One of the parameters of such a distribution is proposed to be used as a measure of image quality. 16 quality measures of the images described in the scientific literature have been selected. It is shown that the scale parameter of the Weibull distribution is a more accurate global quality measure for the set of local estimates than the mean value. A number of experiments have been carried out to confirm the correctness of such an estimate and its correlation with visual estimates of image quality. Such estimates are very important for a) quality assessment of automatically generated photographs, b) selection of parameters for enhancement-oriented image transformations, such as brightness changes, compression of the dynamic range of brightness, conversion to the grayscale representation, and others.

Keywords

• image quality assessment
• no-reference measure
• gamma correction
• linear change of brightness
• contrast
• focal length
• blur
• weibull distribution